1
When shifting the blame is justified: The role of perceptions
of responsibility for negative experiences in the evaluation of
defensive responses to online reviews.
Anna van de Riet
s2911000
June 20, 2016
Supervisor: dr. J.A. Voerman
2
When shifting the blame is justified: The role of perceptions of responsibility for
negative experiences in the evaluation of defensive responses to online reviews.
University of Groningen
Faculty of Economics & Business
MSC Marketing Management
Author: Anna van de Riet
Date: June 20, 2016
Adress: W.A. Scholtenstraat 2-6 9712 KW Groningen
Phone number: 0031 624833148
E-mail adress: annavanderiet@hotmail.com
Student number: 2911000
First supervisor: dr. J.A. Voerman
3
Management summary
This study examines the relationship between the attribution of responsibility for a negative experience in an online review and the satisfaction with a defensive response from a
company. Consumers may write negative online reviews about utilitarian or hedonic products, and provide the reader with arguments based on concrete product related characteristics or subjective reviewer related characteristics. Based on this, a reader possibly perceives that the responsibility for the negative event lies with the company that sold the product, or with the reviewer himself. When a company responds defensively to a negative review, the attribution of responsibility can determine how this response will be evaluated. This research gives valuable insights into how product type and type of arguments influence the attribution of responsibility for negative experiences, and how this in turn affects the satisfaction with defensive responses from companies.
4
Table of contents
Chapter 1: Introduction 1
1.1 Online customer reviews 1
1.2 Negative online customer reviews 1
1.3 Responses to NCR’s and their evaluation 2
1.3.1 Response strategies 2
1.3.2 Evaluation of responses 3
1.4 Attribution theory 4
1.5 Factors influencing attribution of responsibility 5
1.5.1 Product type 5
1.5.2 Type of arguments 5
1.6 Problem statement and research questions 6
1.7 Academic and managerial relevance 6
1.8 Structure of thesis 7
Chapter 2: Theoretical framework 8
2.1 Attribution of responsibility and the evaluation of defensive responses 8
2.2 Attribution based on product type 9
2.2.1 Hedonic and utilitarian products 9
2.3 Attribution based on type of arguments provided 10
2.3.1 Product related arguments 11
2.3.2 Reviewer related arguments 11
2.4 Interaction between type of arguments and attribution based on product 12
2.5 Consumer skepticism 12
2.5.1 Skepticism and attribution of responsibility 13 2.5.2 Skepticism and satisfaction with the defensive response 13
5
3.4.2 Variables and scales 19
3.5 Demographics 23
3.6 Plan of analysis 24
3.6.1 Data preparation 25
3.6.2 Analyses 25
3.7 Results manipulation checks 26
3.7.1 Product type 26
3.7.2 Arguments available 27
3.7.3 Type of arguments 27
3.7.4 Defensive response 27
Chapter 4: Results 29
4.1 Outcomes analysis of variance 29
4.1.1 Main effects 29
4.1.2 Interaction effects 31
4.2 Outcomes regression analysis 32
4.2.1 Regression model with DV = attribution of responsibility 32 4.2.2 Regression model with DV = satisfaction with response 33
4.3 Mediation analysis 34
4.3.1 Explanation of the model 35
4.3.2 Outcomes 36
4.4 Discussion of hypotheses 38
4.4.1 Main effect of product type on attribution of responsibility 38 4.4.2 Main effect of type of arguments on attribution of responsibility 39 4.4.3 Interaction effect of product type and type of arguments on attribution 39 of responsibility
4.4.4 Effect of skepticism on attribution of responsibility 39 4.4.5 Main effect attribution of responsibility on satisfaction with response 40 4.4.6 Effect of skepticism on satisfaction with response 40
Chapter 5: Conclusion 42
5.1 Answering the research question 42
5.2 General discussion 43
5.3 Managerial implications 44
6
References 46
Appendices 50
Appendix A: Survey 50
Appendix B: Reliability analysis 59
Appendix C: Factor Analysis 61
Appendix D: Statistical tests demographics on dependent variables 64
Appendix E: Randomization checks 67
Appendix F: Manipulation checks 69
Appendix G: Analysis of Variance 73
Appendix H: Regression analysis 75
1
Chapter 1: Introduction
1.1 Online customer reviews
In today’s society, people are more and more connected with each other. Customers can provide other customers with word of mouth information about products, services, brands or companies. When such information is communicated via the internet, it is called electronic word of mouth (eWOM) (Babic et. al., 2015). EWOM is important in the consumer decision-making process, with an increasing number of people trusting this form of communication over traditional media (You et. al., 2015). For example, when a consumer is looking at Bol.com for a book, he is offered ratings and comments by fellow consumers in addition to the information provided by the seller. Opinions of consumers are widely and easily
accessible to other consumers on the internet, while the writer of the information only has to interact with a computer. This makes the barrier for consumers to post and read reviews very low. Marketers also recognize the value of eWOM as a source of information and enable and encourage consumers to post reviews about their products (Sen & Lerman, 2007). For
instance, when buying an item from an online retailer such as Amazon or Bol.com, consumers are sent e-mails afterwards asking them to write a review about the product as this may help others. These reviews, also called online customer reviews (OCR’s), have become an important aspect in judging a product or services’ quality nowadays (Ullrich & Brunner, 2015). According to Zhu & Zang (2010, p133) “many people believe that online customer reviews are a good proxy for overall word of mouth and can also influence consumers’ decisions”. This means that people trust more and more on information provided by others before making a purchase decision. Unlike information given by a seller, OCR’s describe product attributes in terms of usage situations and measure the product performance from a users’ perspective (Park et. al., 2007). A product review is typically written to either recommend a product or discourage others from buying it (Sen & Lerman, 2007). Positive and negative reviews can be simultaneously presented together from various sources at the same online platform (Lee et. al., 2008).
1.2 Negative online customer reviews
2 has certain expectations about it. When reality meets or exceeds these expectations,
consumers will be satisfied. However, when the reality does not match up to ones’ expectations about a product, the consumer will be dissatisfied. Consumers with high expectations are more easily dissatisfied than consumers with low expectations about something (Kamins & Assael, 1987). Unsatisfied consumers might feel the need to warn others about a product and post a negative review (NCR). A NCR could make a reader believe that the quality of a product is low, thus reducing purchase intentions (Skowronski &
Carlston, 1987). According to Sen & Lerman (2007), this makes negative information more helpful to consumers in distinguishing good and bad products and it is given greater weight than positive information. Because the social environment of a consumer contains a greater number of positive than negative cues, the negative cues attract more attention. NCR’s therefore tend to have a greater impact on purchase intention than positive reviews
(Weinberger & Dillon, 1980, in Sen & Lerman, 2007). This study is focused only on NCR’s about products, because of their concrete attributes and tangibility. Companies cannot control the negative information that is written about them or their products, but they can control how they react. As consumers will be more likely to consider negative e-WOM reviews than positive reviews for their decision-making, a satisfactory response from a firm to online complaints can be crucial in terms of customer retention and corporate reputation (Breitsohl et. al., 2010; Sen & Lerman, 2007).
1.3 Responses to NCR’s and their evaluation
It can be critical for companies to respond proactively to online complaints, because it may prevent further attacks from other consumers (Lee & Song, 2010). To protect or improve the reputation of a firm it is necessary to handle NCR’s with the appropriate strategy (Davidow, 2003, in Lee & Song, 2010). The perception that potential customers have towards a company is very important in its marketing activities. Consumers read complaints and negative
information from others online and interpret and perceive them for their own needs (Swanson, 1987, in Lee & Song, 2010). Therefore, it is important that these readers will evaluate
responses to NCR’s positively, as they are potential future customers. 1.3.1 Response strategies
3 response from a brand or consumer than no response at all. Responses, as opposed to silence, positively influence the brand evaluations of consumers who were exposed to negative
reviews. So, responding to a NCR is the best option for companies. Then the question remains how they should respond.
Lee & Song (2010) state that proactive actions from a company on a negative event help restore their positive image. In their research, they distinguish between proactive
(accommodative) and defensive reactions. In a defensive response, the organizational interest is put first. Defensive strategies deny responsibility for a negative event and possibly shift the blame to others (Lee & Song, 2010). A strategy like this can also be described as using excuses. Excuses can be defined as “explanations that remove the organization from
responsibility for its predicament” (Conlon & Murray, 1996, p1042). An excuse is used when a company admits that an act was done but denies being personally responsible for it (Bolkan & Daly, 2009). Accommodative responses are any form of apology and/or corrective actions, such as compensation. With an apology, a firm regrets that consumers have been
inconvenienced and admits that a mistake was made. Customers often expect this kind of reaction as an acknowledgement that they have been wronged (Bolkan & Daly, 2009). A corrective action is a form of tangible compensation for a negative event. Receiving
compensation enhances satisfaction with organizational responses and willingness to remain a customer of the firm in the future (Conlon & Murray, 1996). Accommodative responses create more satisfaction than defensive responses, although the company is regarded as being at fault. With an accommodative response, the complainers’ concerns are put first. Companies who use accommodative responses show willingness to publicly accept responsibility for the failure that occurred. The acceptation of responsibility facilitates consumers’ trust in a firm, which affects their evaluation positively (Lee & Song, 2010). Defensive strategies might be useful when it is hard to identify who the source of the problem is. However, consumers may feel easily disappointed by company responses like shifting blame to the consumer (Lee & Song, 2010).
1.3.2 Evaluation of responses
4 satisfaction after complaints. As defensive responses lead to more dissatisfaction than
accommodative responses, they are much less used by companies (Einwiller & Steilen, 2015). The results of Einwiller & Steilen (2015) and Lee & Song (2010) show that accommodative strategies have a stronger overall impact on consumers’ evaluation of the firm and satisfaction with a response than defensive strategies. This suggests that consumers value any form of accommodative actions of a company in an online context, especially when corrective actions are involved. So, in general, an accommodative response is a better option for companies than a defensive response. However, evidence by Jin et. al. (2014) suggests that defensive
reactions are acceptable to the public in a crisis situation where the origin is perceived to be external, so outside the fault of the company. Whether the cause for a negative event is perceived to be in or outside the fault of a company is dependent on the attribution of responsibility (Folkes, 1984). The attribution theory will be explained in more detail in the next paragraph.
This research is focused on situations in which defensive responses might be evaluated positively by readers of NCR’s, and the purpose is to find out when defensive responses from organizations create satisfaction with readers of negative reviews. It might not always be necessary to take responsibility for a negative experience from a customer, because the cause may be outside the fault of the company. Since it is possible that taking responsibility results in loss of reputation, and corrective actions cost the organization money, a defensive response could sometimes be an attractive alternative to an accommodative response.
1.4 Attribution theory
Readers of online complaints try to identify who is responsible for the negative event and/or what caused the problem (Lee & Song, 2010). This may in turn affect their perception of the company that is involved. Sen & Lerman (2007) use the attribution theory paradigm to
5 responsibility for the experience to the reviewer or to the company that provided the product. The subsequent attitudes of the reader of a NCR can be affected by the type of attribution.
1.5 Factors influencing attribution of responsibility
Whether a reader of a NCR assigns responsibility for a negative experience to the company or to the writer of a review may depend on certain factors. As mentioned in paragraph 1.3, the evaluation of organizational responses to NCR’s can be influenced by the attribution of responsibility. Hareli (2005, in Bolkan & Daly, 2009) says that comparing different types of organizational responses to consumer complaints is misguided, because distinctions cannot be made purely on their form, context must be considered. In this research, the product type involved and the type of arguments provided are considered as context factors that may influence the satisfaction with defensive organizational responses. This is because both the product type and the type of arguments can have an influence on the attribution of
responsibility. 1.5.1 Product type
A distinction in product types can be made between hedonic and utilitarian products (Dhar & Wertenbroch, 2000). Hirschman & Holbrook (1982) describe hedonic products (e.g., music, art, movies) as products that satisfy emotional wants, and are consumed to create sensual pleasure, fantasy and fun. They deliver an affective and sensory experience. Utilitarian products (e.g., dishwashers, consumer durables) provide a solution to completing a functional or practical task, and their consumption is cognitively driven, instrumental and goal-oriented (Strahilevitz & Myers, 1998, in Dhar & Wertenbroch, 2000). Consumers generally assign greater importance to hedonic attributes of consumption when evaluating hedonic products, as opposed to concrete attributes in evaluating utilitarian products. As a consequence, negative reviews about hedonic products are perceived as less helpful to readers because the expressed opinions are subjective, based on the reviewer himself. Utilitarian NCR’s are seen as more objective and focused on the product characteristics (Sen & Lerman, 2007).
1.5.2 Type of arguments
6 written (Ghose & Ipeirotis, 2011; Park et. al., 2007). Thus, the type of arguments provided in a review may also affect how readers attribute responsibility for a negative experience. If readers make the attribution that a NCR is based on product reasons, they will perceive it as legitimate, believable and useful. On the other hand, if a reader believes that a review is based on reasons related to the reviewer, he will not find it helpful (Lee et. al., 2008). The type of arguments provided in the review (product- or reviewer related) can exist for both utilitarian and hedonic products. This may include information about the packaging or material of a hedonic product or the usage situation of a utilitarian product. Therefore, the type of arguments is included as a separate variable.
1.6 Problem statement and research questions
The hedonic or utilitarian product type and the type of arguments provided in a review might both have an influence on the way readers blame failures in NCR’s on the seller or the reviewer. The perceived responsibility for a negative experience in a review may cause defensive responses to be evaluated positively in certain situations. This thesis is focused on the influence of product type and type of arguments on the satisfaction with defensive
responses from companies to NCR’s. Having insight in the product characteristics and type of arguments that may cause defensive responses to be evaluated positively can give more clarity about situations in which these responses are suitable.
The following main research question is conducted:
“In what way does the attribution of responsibility for a negative experience, by means of product type and type of arguments, influence the evaluation of defensive responses to negative online customer reviews?”
This main question can be answered by researching the following questions: 1. How does product type influence attribution of responsibility?
2. How does the type of arguments provided influence attribution of responsibility? 3. How does the attribution of responsibility affect satisfaction with defensive responses? 4. In which situations is a defensive response appropriate?
1.7 Academic and managerial relevance
7 be taken into account when deciding on response strategies. Without further research in this area it remains uncertain whether accommodative responses are preferred in all situations because of the responsibility and/or the corrective actions that are taken. The product type and the type of arguments involved in a review can influence who is perceived to be at fault for a negative event in a NCR. The blame that is attributed to a seller or reviewer might in turn influence the success of a certain response strategy. It is interesting to know how the product type and type of arguments influence the evaluation of defensive responses because it can generate new insights in the situations in which they create satisfaction. For managers this might implicate that it is not always necessary for them to take responsibility for a failure and spend money on corrective actions. In some situations the company could not be perceived as the one to blame and the response strategy should be adjusted accordingly. When defensive reactions are satisfying in some situations, knowledge about these results could contribute to managerial decisions about response strategies to NCR’s. The goal of this research is
consequently to gain insight in the influence of hedonic/utilitarian product characteristics and product/reviewer related arguments on the evaluation of defensive organizational responses.
1.8 Structure of thesis
8
Chapter 2: Theoretical framework
This chapter starts by explaining how the attribution of responsibility for a negative
experience in a review may influence the evaluation of defensive responses from companies. As explained in chapter one, a defensive response consists of denial, attack or shifting the blame. The company admits that a negative event has happened, but attempts to minimize the organization’s responsibility for it (Lee & Song, 2010). The independent variables, i.e. product type and type of arguments and the mediating variable attribution of responsibility will be described in depth using relevant academic literature. Insight will be given into their relationship with the dependent variable, satisfaction with defensive organizational responses. After this, attribution of responsibility is explained by means of the two independent
variables, product type and type of arguments. Following, a possible interaction effect between the type of arguments and the attribution based on product type is explained. Lastly, skepticism as a consumer characteristic will be discussed because this possibly influences the attribution of responsibility and/or the satisfaction with the response. Each section will be concluded with subsequent hypotheses. The conceptual model is shown at the end of this chapter, which represents a graphical depiction of the relationship between the variables. 2.1 Attribution of responsibility and the evaluation of defensive responses
According to the attribution theory, people are rational information processors whose actions are influenced by their causal inferences (Folkes, 1984). When a certain event occurs, people seek causes for it in a variety of domains. If a product fails, they try to determine why it failed. In consumer complaining behavior, attribution theory predicts that consumers’ reactions will be influenced by what they perceive is the cause for a product failure. Responsibility can be used as a form of attitudinal judgment on causes for negative events (Folkes, 1984; Lee & Song, 2010). In this study, attribution is defined as the opinion that a reader generates to define who is responsible for an event. The vast majority of complainants are convinced that the company is responsible for the problem (Einwiller & Steilen, 2015). The reader of a NCR does not have to agree with this. Readers may either feel the company or the writer of the review is responsible for the negative event. Folkes (1984) mentions locus of control as a dimension of attribution. This means whether the cause of a failure has something to do with the consumer or the producer or distributor of the product. Locus of control
9 exchange upsets the relationship between the buyer and seller and the consumer has been wronged. This relationship needs to be repaired, which can be done by apologizing or offering corrective actions. So, accommodative responses are favorable and expected when the reader thinks the firm is responsible (Einwiller & Steilen, 2015; Lee & Song, 2010). Defensive responses are often disliked because the company does not take responsibility and tries to shift the blame. However, they might be appropriate in situations where the origin of a product failure is perceived to be with the reviewer (Jin et. al., 2014). When the consumer is responsible for the failure, the firm does not owe a consumer anything according to Folkes (1984). This is when a defensive response might also lead to satisfaction with readers of NCR’s. The following hypothesis is proposed:
H1: The satisfaction with defensive responses by companies to NCR’s is higher when the responsibility for a product failure is assigned more to the consumer as opposed to the company.
2.2 Attribution based on product type
The perception of responsibility may depend on the product type that is involved in the
review. When a review contains no information but only a negative opinion about a product, it is assumed that the attribution of responsibility is dependent on the characteristics of the product type involved. A distinction in product type can be made between hedonic and utilitarian products (Dhar & Wertenbroch, 2000). The product types are different in why they are consumed and how people evaluate them.
2.2.1 Hedonic and utilitarian products
10 were guided by personal reasons (Sen & Lerman, 2007).
For utilitarian products, the concrete attributes are most important in the evaluation (Sen & Lerman, 2007). Consumers’ primary concern is with the immediate consequences of consumption (Batra & Ahtola, 2001, in Sen & Lerman, 2007). Negative experiences with a product will impact the utility of that product for the consumer. Since utilitarian consumption is concerned with maximizing utility, negative information about a product will be weighed more heavily when evaluating utilitarian products. Utility maximization is characterized by tangible and seemingly objective criteria (Sen & Lerman, 2007). Therefore, readers feel comfortable relying on the information given in NCR’s of utilitarian products.
Cervellon & Carey (2014) also found that reading negative reviews influences consumers’ evaluations more for utilitarian than for hedonic products. Consumers are more likely to believe that the reviewer is knowledgeable and trustworthy. The reader generally assumes that the motive for writing the review was to accurately inform others about a product, and that the review is based on the reviewer’s true experiences. As a consequence, NCR’s are more
helpful to consumers of utilitarian products than for consumers of hedonic products (Sen & Lerman, 2007).
Hedonic products are evaluated on the sensory experience they bring, which is subjective and often dependent on personal taste. A negative experience is likely to be attributed to
characteristics and personal situations of the writer, and in that case the company is probably not seen as the one to blame. The responsibility for the failure of the product to satisfy the buyer is more likely to be assigned to the reviewer for hedonic products. Evaluation of utilitarian products is based on concrete attributes and seemingly objective criteria (Sen & Lerman, 2007). Failure can be attributed to the product itself, therefore likely regarded as being the companies’ fault. The following hypothesis is proposed:
H2: When a product is more hedonic (utilitarian) the responsibility for a negative experience in a NCR is more likely to be assigned to the consumer (company) as opposed to the company (consumer).
2.3 Attribution based on type of arguments provided
11 attributes of a product itself, or about the personal situation of the writer of the review.
Suggested is that the attribution of responsibility may also depend on the type of arguments that is provided in a NCR. Readers of NCR’s trust that the opinions expressed are either based on product related reasons or on subjective reviewer related reasons. This will determine how useful a review is to them (Sen & Lerman, 2007). As mentioned in chapter one, reviews about hedonic products can as well provide concrete product related arguments, and reviews about utilitarian products can contain subjective reviewer related opinions. Therefore, the type of arguments is considered as a separate variable in this research.
2.3.1 Product related arguments
Reviews that are based on product related arguments contain information about concrete attributes of a product (Sen & Lerman, 2007). The reviewer writes “objective statements” that portray factual data about product features. Characteristics of a product are listed and a
product description is given that, in case of a NCR, rejects the description written by the seller (Ghose & Ipeirotis, 2011). The arguments are relevant to evaluate the product for any reader and are reliable (Lee et. al., 2008). Readers consider reviews that are based on product related arguments useful, because they seem legitimate and believable to them (Sen & Lerman, 2007). The following text is an example of product related arguments in a NCR:
“This product has very limited battery life. It didn’t come with an AC power source. It also has no hold button, which means I have to take out the batteries when I’m not listening. Sometimes, it makes a really high pitched buzz in the earphones” (Lee et. al., 2008, p343).
2.3.2 Reviewer related arguments
Reviewers can also write down “subjective opinions” that portray the reviewers’ emotions about product features. The arguments are sentimental and the reviewer gives a very personal description of a product. The given information does typically not appear in any official description of the product (Ghose & Ipeirotis, 2011). Reviews that are based on reviewer related arguments contain information that is attributable to personal reasons of the writer and his situation. Examples are specific user situations and personal opinions that are not
12
“I got this product four weeks ago. I purchased it for my son for our trip to Disney. He loved it but after one week, he didn’t play with it anymore. Hmmm... This product is not what he wants. Mistake! I shouldn’t have chosen it” (Lee et. al., 2008, p343).
Product related arguments are trustworthy for readers because they are about concrete
attributes (Sen & Lerman, 2007). Because product related arguments are about the failure in a product itself, the responsibility is likely to be attributed to the company. Reviewer related arguments are seen as unreliable because they are about personal reasons (Sen & Lerman, 2007). When the arguments provided are reviewer related, the readers will likely perceive the writer as being responsible for the negative experience. The following hypothesis is proposed:
H3: Responsibility for a negative experience in a NCR is more likely to be assigned to the company (consumer) as opposed to the consumer (company) when provided with product related arguments (reviewer related arguments).
2.4 Interaction between type of arguments and attribution based on product type
When only looking at product characteristics, it is proposed that reviews about hedonic products will result in assigning responsibility to the consumer. This is because hedonic products are evaluated on subjective criteria (Sen & Lerman, 2007). Utilitarian products are evaluated on their concrete characteristics, making it more likely to assign responsibility to the company. However, when there are arguments added in a review, it is assumed that this is more important than product characteristics. So, when adding product- or reviewer related arguments, the responsibility assigned depends on the type of arguments provided and no longer on the product type. This means that there is possibly an interaction effect between type of arguments and the relationship between product type and attribution of responsibility. Product type may only influence attribution of responsibility when no arguments are provided in a review. Reviews containing reviewer related arguments, for both utilitarian and hedonic products, might result in assigning responsibility to the consumer and the other way around. The following hypothesis is proposed:
H4: When product- or reviewer related arguments are provided in a review, the product type does not influence attribution of responsibility.
2.5 Consumer skepticism
13 significant results in this research are found, it is expected that they are due to the
manipulations of the independent variables. There might however be other factors that
influence the attribution of responsibility or the satisfaction with a defensive response, such as personal characteristics of the respondent.A consumer characteristic that can possibly
influence the tendency to attribute failure to a consumer or company when reading NCR’s is skepticism. Skepticism is defined by Skarmeas & Leonidou (2013) as the tendency to disbelieve, doubt and question information from others. A consumer can also be skeptic towards actions by organizations (Skarmeas & Leonidou, 2013). This may influence how a response from a company is evaluated. The two forms of skepticism as consumer
characteristics are used as control variables in this research. 2.5.1 Skepticism and attribution of responsibility
When people are skeptic towards information provided by others, they might easily attribute responsibility for a failure to a consumer when reading a NCR. If people are skeptic towards organizations however, they might be more inclined to assign responsibility to a company. The following hypothesis is proposed:
H5: Consumers who are skeptic towards other people (organizations) are more likely to attribute responsibility for a negative experience in a NCR to the consumer (company) as opposed to the company (consumer).
2.5.2 Skepticism and satisfaction with the defensive response
When consumers are skeptic towards organizations, they might doubt or disbelieve any information provided by companies. Therefore, the satisfaction with a defensive response from a company might be lower than for consumers who are not skeptic towards
organizations. The following hypothesis is conducted:
H6: Consumers who are skeptic towards organizations are more likely to evaluate a defensive response from a company negatively.
2.6 Conceptual model
On account of the previous theories, the conceptual model depicted in figure 1 is drawn. The two independent variables are represented: product type and type of arguments. These
14 responsibility, so the effect of the independent variables goes through the mediating variable. The control variable, consumer skepticism, is also expected to influence the attribution of responsibility and the satisfaction with the response. The arrows represent hypotheses about the variables. The following chapter will present the methodology and measures that will be used in this research.
Figure 1: Conceptual model
Hypothesis 1 The satisfaction with defensive responses by companies to NCR’s is higher when the responsibility for a product failure is assigned more to the
consumer as opposed to the company
Hypothesis 2 When a product is more hedonic (utilitarian) the responsibility for a
negative experience in a NCR is more likely to be assigned to the consumer (company) as opposed to the company (consumer)
Hypothesis 3 Responsibility for a negative experience in a NCR is more likely to be assigned to the company (consumer) as opposed to the consumer
(company) when provided with product related arguments (reviewer related arguments)
Hypothesis 4 When product- or reviewer related arguments are provided in a review, the product type does not influence attribution of responsibility
Hypothesis 5 Consumers who are skeptic towards other people (organizations) are more likely to attribute responsibility for a negative experience in a NCR to the consumer (company) as opposed to the company (consumer)
Hypothesis 6 Consumers who are skeptic towards organizations are more likely to evaluate a defensive response from a company negatively
15
Chapter 3: Methodology
In this chapter, the research design of this study will be explained. First, the chosen
experimental design will be justified, and the data collection procedure will be discussed in detail. The operationalization of the variables is presented and the plan of analysis will be explained. The results of the manipulation checks and an overview of the demographics of the sample are given as well.
3.1 Research design
To gain more insight in the relationships between the variables in this research, an
16
No arguments Product related arguments
Reviewer related arguments Hedonic Condition 1 Condition 2 Condition 3
Utilitarian Condition 4 Condition 5 Condition 6 Table 2: Conditions
3.2 Survey
The data will be gathered by doing an online survey. It is important that the respondents all answer the same questions, regardless of the condition they get to see. The questions are about judging NCR’s about perception of responsibility and satisfaction with defensive responses on a certain scale. When this scale is the same in all cases, the conditions are easily compared with each other. A survey with closed questions is suitable for this comparison. When asking respondents to fill out a survey, there is no distraction or influencing by a researcher who is present (Vennix, 2011). The survey is chosen to be distributed online. This way many different respondents can be reached and they can answer the questions on a moment that is convenient for them.
3.3 Procedure
The target group of this experiment is adults. This is because this target group is most likely to consist of typical people who read online customer reviews (Marchant, 2015). The definition of adults is people in the age category of 18-65 years old (Encyclo, 2016). The survey will be spread online through e-mail and different social media channels. The
respondents will first read a short introduction about the background of this research project and the amount of time it will take to finish the survey. It will be highlighted that the
participants will remain anonymous. After starting the survey, the respondent will see some pictures of either a hedonic (DVD) or utilitarian (Vacuum cleaner) product. A negative review about this product is shown. This review contains either no arguments, or product- or
17 asked to judge the product on the extent to which it is hedonic or utilitarian through a few items. The level of skepticism towards others and organizations of the participant will also be measured through certain questions. At the end some demographic questions about the gender, age and educational level of the respondent will be asked. The participants will be thanked for their cooperation.
3.4 Operationalization
3.4.1 Manipulations
The independent variables in this research are manipulated by the researcher, which is done based on literature. The hedonic product that is shown in the survey is a movie (DVD), as Hirschman & Holbrook (1982) name movies as one of the typical examples of hedonic products. Also, it is used in the research design of Sen & Lerman (2007). The picture is of a DVD called ‘A perfect stranger’. This picture is chosen because there are both a man and woman on the cover, and as a thriller it is not a typical movie for either men or women. The utilitarian product is chosen to be a vacuum cleaner, named a typical example of a utilitarian product by Dhar & Wertenbroch (2000). Sen & Lerman (2007) also state that typical
utilitarian products are consumer durables. It will be explained in the survey that the reviewer bought the product at an electronics retailer.
Figure 2: Example hedonic and utilitarian product
18 that are based on personal situations of the writer. This is how and Lee et. al. (2008) and Sen & Lerman (2007) describe reviewer related information, in this case arguments, in reviews. The reviews containing product related arguments describe concrete product attributes that are objective and relevant to evaluate a product. Lee et. al. (2008) and Sen & Lerman (2007) define product related arguments this way. All the reviews end with the phrase “would not recommend” to highlight the negative tone. Lee & Song (2010) describe the main
characteristics of a defensive organizational response as denying responsibility for a negative event and shifting the responsibility to another party. The most important differences with an accommodative response are that a defensive response does not contain an apology and that no responsibility is taken. The response to the NCR shown in this research denies
responsibility for the negative event by saying that they do not feel that they are responsible for the negative experience. No apology is given. The responsibility is also shifted to the consumer by stating that it is a well sold product that other customers are enthusiastic about.
Reviews
No arguments “I did not like this product at all. Would not recommend.” Hedonic -
Reviewer arguments
“Found this movie not inspiring, I did not relate to the characters. Started to watch it on my free Sunday evening but halfway I turned it off
because I would rather finish my book. Would not recommend.” Utilitarian –
Reviewer arguments
“Bought this vacuum cleaner for cleaning the upstairs bedrooms but I have back problems and it is too heavy for me to take up the stairs. I never use it. Would not recommend.”
Hedonic – Product arguments
“When I came home, I saw that the box was broken. Also, the disc had scratches on it which made it skip over every 5 minutes. Would not recommend.”
Utilitarian - Product arguments
“The cleaning performance of this vacuum cleaner is very poor. It makes a lot of noise and the power cord is very short. Would not recommend.”
Defensive response
“We find it unfortunate to hear that the product does not meet your expectations. However, we do not think that we are responsible for your negative experience. This is a well-sold product, and many customers are enthusiastic about it. We hope to be of service to you in the future.”
19 3.4.2 Variables and scales
The constructs that are used in this research are represented by certain variables. These variables are measured by items that form scales, of which an overview can be found in table 4. To measure the reliability of the scales it is necessary to check to what extent the items correlate with each other. This correlation between the items was calculated with a reliability analysis. Reliability means that a measurement instrument consistently measures a construct it intends to measure (Vennix, 2010). The general rule is that a scale with a Cronbach’s alpha of less than .60 is unreliable, and that an alpha above .80 is very reliable (Field, 2013). For certain scales, factor analysis was performed to see whether the items all load on the same construct and could be used to measure the intended variable. The output of the factor and reliability analyses can be found in Appendix B and C.
Concept Item Scale α
Dependent variable Satisfaction with
defensive organizational response
Source:
Eroglu & Machleit, 1990
1 Unsatisfied – Satisfied 2 Displeased - Pleased 3 Unpleasant - Pleasant
4 I do not like it at all – I like it very much 7 point semantic differential .964 Mediating variable Perceived organizational responsibility for the negative experience Source: Lee (2004)
1 To what degree should the retailer be blamed for the negative experience 2 How much responsibility should the retailer bear for the negative
experience 7 point Semantic differential 1 = Not at all 7 = Totally .894
Independent variables – manipulation check Product type (hedonic)
Source:
Voss, Spangenberg, Grohmann, 2003
1 Not fun – Fun 2 Dull – Exciting
3 Not delightful – Delightful 4 Not thrilling – Thrilling 5 Unenjoyable – Enjoyable
7 point Semantic differential
.935
Product type (utilitarian) 1 Not effective – Effective 2 Not helpful – Helpful
7 point Semantic
20 Source:
Voss, Spangenberg, Grohmann, 2003
3 Not functional - Functional 4 Not necessary - Necessary 5 Not practical – Practical
differential
Arguments available 1 This review contains arguments that support the writer’s opinion about the product 7 point Semantic differential 1 = Strongly disagree 2 = Strongly agree Type of arguments 1 There are no arguments about the
concrete characteristics of the product in the review
2 The arguments in the review are relevant to evaluate the product 3 The arguments in the review are about the personal situation of the writer 7 point Semantic differential 1 = Strongly disagree 7 = Strongly agree .707
Dependent variable – manipulation check Defensive organizational
response Source: Xia (2013)
1 The retailer admits being at fault in the response
2 The retailer’s response is very defensive 7 point Semantic differential 1 = Strongly disagree 7 = Strongly agree Control variables Skepticism towards others Source: Hurtt (2010)
1 I often accept other people’s explanations without further thought 2 I often reject statements
unless I have proof that they are true 3 I tend to immediately accept what other people tell me.
4 I frequently question things that I see or hear.
5 It is easy for other people to
21 convince me
Skepticism towards organizations
Source: Skarmeas & Leonidou (2013)
1 Retailers generally act socially responsible
2 Retailers are generally concerned to improve the well-being of consumers 3 Retailers generally follow high ethical standards 7 point Semantic differential 1 = Strongly disagree 7 = Strongly agree .699
Table 4: Operationalization table
Satisfaction with the defensive organizational response
Satisfaction with defensive organizational response is the dependent variable in this thesis. Satisfaction is the fulfillment of one’s wishes, expectations, needs or the pleasure derived from this (Oxford dictionaries, 2016). In this thesis, the operational definition of the variable is the extent to which a reader of a NCR is satisfied with the defensive organizational
response. Four items are used to measure this variable, derived from the Marketing scales handbook by Bruner ǁǁ & Hensel (1998). The reliability analysis shows that the scale has sufficient reliability (α= .964).
Attribution of responsibility
Attribution of responsibility is the mediating variable in this research. Attributions are the causes inferred for product failure (Folkes, 1984). In this research, attribution of responsibility is defined as the opinion that a reader generates to define who is responsible for an event. Here, it is chosen to measure attribution to the retailer instead of a manufacturer of a product. This is because people buy products at retailers and turn to them if they have complaints. It is assumed that consumers hold the one that sold them a product, the retailer, responsible for any possible failures. To measure this variable, respondents are asked to judge a review based on two items. The original scale is from Lee (2004) and measures perceived organizational responsibility for a crisis. Factor analysis showed that the two items measure one construct and the reliability of the scale is high (α= .894).
Product type
22 hedonic or utilitarian. There is a separate scale for both conditions, using a 7 point semantic differential. Both scales consist of five items. Reliability analysis shows that both scales are reliable (α= .935 & .949).
Type of arguments
The type of arguments provided is also a variable that is manipulated by the researcher. It consists of three levels: no arguments, product related arguments and reviewer related arguments. Product related arguments are about concrete attributes of a product and are relevant to evaluate it. Reviewer related arguments are attributable to personal reasons of the writer of a review and are usually not seen as relevant (Lee et. al., 2008; Sen & Lerman, 2007). The operational definition in this research is the extent to which a reader perceives the arguments provided in a NCR to be related to the product or the reviewer. The participants will answer a question in the survey to test if they perceive that the review contains arguments or not. When there are arguments provided in the review, the respondents will also answer a few questions regarding the characteristics of these arguments to test whether the
manipulation was successful. The scale that measures whether arguments are product- or reviewer related was constructed based on characteristics of the different types of arguments, found in the literature. It contains three items, of which the second one is reversed coded. Factor analysis shows that the items load on one construct and the reliability analysis shows that the scale has a high reliability (α= .707).
Defensive response
The dependent variable in this research is about defensive responses from organizations to NCR’s. Defensive responses put the organizational interest first and deny responsibility for a negative event. The blame is shifted away from the company (Lee & Song, 2010). To check if the manipulation of the defensive response type has worked two questions derived by Xia (2013) are asked in the survey. The original scale uses the word ‘brand’ which is replaced by ‘the retailer’. The reliability analysis showed that the two items of the scale do not measure the same construct according to the respondents (α= .090). Therefore, the two items will be analyzed separately.
Skepticism towards others
23 professional skepticism by Hurtt (2010) is used in this thesis. This is originally a 30 item scale, of which five items are chosen that resemble skepticism toward other people. A low score in this scale indicates that the respondent has a high level of skepticism and a high score indicates a low level of skepticism. Items 2 and 4 are reversed to check if the respondents pay attention to the questions. Reliability analysis shows that the scale is reliable (α= .741).
Skepticism towards organizations
Skepticism towards organizations can be explained as the extent to which consumers doubt that companies live up to their professed standards (Skarmeas & Leonidou, 2013). This is measured by a scale from Skarmeas & Leonidou (2013) about CSR skepticism. The original scale consists of four items, of which three are chosen to be a good fit for this research. A high score indicates a low level of skepticism towards organizations and a low score indicates a high level of skepticism. Factor analysis showed that the items clearly load on one construct. Additionally, reliability analysis was performed and the scale is reliable (α= .699).
3.5 Demographics
The sample consists of 197 people from the intended research population of adults. Due to incomplete data of some of the respondents, six surveys needed to be excluded. Because of this exclusion, some of the conditions received fewer than 30 respondents. 190 people answered the question about their gender, of which 132 were female (69.5 %) and 58 male (30.5 %). The age of the respondents varies from 18 to 62 years old, with an average age of 27.2 and the most frequent age being 23. The population is well-educated, with respectively 49 and 25 percent having a Bachelor’s and a Master’s degree.
Condition Mean age Gender Male Female N 1 24 20.7 % 79.3 % 29 2 24 35.5 % 64.5 % 34 3 26 38.5 % 61.5 % 28 4 30 27.3 % 72.7 % 34 5 27 28.9 % 71.1 % 39 6 29 33.3 % 66.7 % 33
24 To see whether there are significant differences between the groups on the independent
variables gender, age and education level on the dependent variable and the mediator, some tests are conducted. The output of these tests can be found in Appendix D. The results show that all outcomes on the dependent variable are insignificant. This means that there are no significant differences between the variables gender, age and education level on attribution of responsibility and satisfaction with the defensive response.
p-value Df F p-value Df F
DV Mediator
Gender .593 1 .287 .689 1 .161
Age .673 1 .178 .159 1 1.999
Education level .807 5 .458 .555 5 .794 Table 6: Significance population on DV and Mediator
To be able to see whether there are systematic differences between the respondents in the separate conditions, a randomization check has been performed. Some analyses have been conducted where the independent variable was the condition and the dependent variable gender, age or education level. The results of the Chi-square tests (p= .733) (p= .523) and the one-way ANOVA for age (F (5)=1.99, p>.05) show that all outcomes are insignificant. This means that there are no significant differences between the groups, and that the assignment of the participants to a condition in a random way was successful (See Appendix E).
Variable Significance Df
Gender .733 5
Age .082 5
Education level .523 25 Table 7: Randomization check
3.6 Plan of analysis
25 3.6.1 Data preparation
The data that were gathered are checked for missing values and invalid responses. The respondents that did not fill in the survey sufficiently or in an incorrect way are filtered out. With the remaining data set, further analyses will be conducted. The construct validity is tested, to see if the collected data fits the intended hypotheses. This has been done with a principal components factor analysis (Malhotra, 2007). Although most of the measurement scales in this research are developed and tested by other academics, some questions are combined or adapted to the situation in this thesis. Therefore, the factor analysis was performed on the items of the manipulation check for the independent variable type of arguments, as well as the mediating variable and the control variables. The underlying constructs are checked and whether the items could be combined into new factors.
Additionally, the Cronbach alpha’s and KMO’s for the variables were computed to test the reliability and validity of the measures and sample. The degree of consistency of the results is measured by conducting a reliability analysis. The distribution of the variables for the analysis has also been checked for normality. The dependent variable was skewed; therefore the square root of the variable has been taken to be able to use it in further analysis. 3.5.2 Analyses
First, a two-way ANOVA test will be performed to get an overview of the first results of the research. Regression analysis will be conducted to test the effect of the independent variables on the mediator and the dependent variable, and the mediator on the dependent variable. Because the independent variables have multiple categories, dummy variables will be made. Two regression models will be produced. The first model uses attribution of responsibility as the dependent variable. In the second model, the dependent variable is satisfaction with the response. The models will be built up, to see if changes in the relationships between variables occur when other variables are added. The first model that will be produced for each
dependent variable only contains the main effects of the independent variables. The regression equation for this model is Y= b0 + b1X1 + b2X2. Then, the interaction effects are added to the model. Two interaction terms will be produced; product type * reviewer arguments and product type * product arguments. The regression equation for the model including the interaction effects is Y= b0 + b1X1 + b2X2 + b3X1X2. In the regression model with
26 In the conceptual model, the effect of the independent variables on the dependent variable goes through a mediating variable. To check if mediation occurs, the Baron and Kenny mediation test will be used. According to Baron & Kenny (1986), mediation only happens when three conditions take place simultaneously:
-The independent variable has a significant effect on the mediator
-The independent variable has a significant effect on the dependent variable -The mediator has a significant effect on the dependent variable
If these three conditions take place, the full model will be tested by doing bootstrap analysis with the PROCESS macro by Hayes (2013). The model that will be used is number 9, because it allows two moderators that influence the relationship between the independent variable and the mediator. The moderators in this case are the two dummy variables for type of arguments, since the model does not allow multiple X variables.
3.7 Results manipulation checks
The type of product, type of arguments and the defensive response are all manipulated by the researcher. To see whether the manipulation of the research conditions was successful, the mean scores are computed and several tests were conducted to analyze whether there were significant differences between the conditions. As expected, the manipulation of the two independent variables and the dependent variable worked as intended. In the following sections, the results per manipulation will be outlined in detail. The SPSS outputs of the statistical tests can be found in Appendix F.
3.7.1 Product type
27
Product type Mean SD p-value
Hedonic 5.01 1.271 .000
Utilitarian 5.72 1.357 .000 Table 8: Manipulation product type
3.7.2 Arguments available
For the arguments available manipulation, a high score on the scale means that arguments were available and a low score that arguments were not available. When looking at the means, the no argument conditions have a lower score than the argument conditions. The ANOVA test (Appendix F) shows that the groups are significantly different from each other. This means that the participants clearly perceived the no arguments conditions to be different from the other conditions were arguments were available.
Condition Mean SD P-value
Arguments available 4.73 1.662 .000 No arguments available 2.69 1.988
Table 9: Manipulation check arguments available 3.7.3 Type of arguments
On the scale for type of arguments, a high score means that the arguments were reviewer related. When looking at the mean scores, the reviewer arguments conditions score indeed higher than the product arguments conditions. An ANOVA test (Appendix F) was performed and this showed that the groups are significantly different from each other. It can be
concluded that the manipulation worked and participants viewed reviewer related arguments as being different from product related arguments.
Condition Mean SD P-value
Reviewer arguments 5.29 1.029 .000 Product arguments 3.63 1.164
Table 10: Manipulation check type of arguments 3.7.4 Defensive response
28 (Appendix F) were conducted to see if the mean scores are different from the average of the scale, 3.5. For both items, the mean score was significantly lower and higher than the average of the scale. This means that all participants indeed perceived the response from the company as defensive.
Variable Mean SD p-value
The retailer admits being at fault in the response
2.01 1.346 .000
The retailer’s response is defensive
5.04 1.630 .000
29
Chapter 4: Results
In this chapter, the results of the research will be discussed. A total of 197 respondents took part in the online survey, of which 191 finished it completely. First, the results of the analysis of variance and the regression analyses are discussed. In addition, the results of the mediation analysis are presented. To conclude, the findings of the analyses are used to discuss the separate hypotheses.
4.1 Outcomes analysis of variance
To start, a two-way ANOVA test was conducted to gain a first understanding of the relationships between the variables. A first analysis was done with the two independent variables product type and type of arguments, and attribution of responsibility as the dependent variable. This test was repeated with satisfaction with the response as the
dependent variable. An overview of the results of these tests can be seen in tables 12 and 13. The full output is placed in Appendix G.
4.1.1 Main effects
The results of the ANOVA-tests show that product type does not significantly influence attribution of responsibility or satisfaction with the response. The type of arguments does have an influence on attribution of responsibility and also on satisfaction with the response. The results of the Post-hoc test (Appendix G) for attribution of responsibility show that reviewer arguments score lower on attribution to the retailer than product or no arguments. Product arguments have the highest score on attribution to the retailer and the no arguments conditions are in between. This means that the responsibility is more assigned to the retailer when product related arguments are provided, and less when reviewer or no arguments are shown. On satisfaction with the response, reviewer arguments score higher than product arguments and no arguments. This means that when presented with reviewer related
30
DV: Attribution of responsibility
Condition Hedonic Utilitarian Total
M M M
Reviewer arguments 1.72 (.872) 2.44 (1.295) 2.11 (1.169)
Product arguments 5.46 (1.233) 4.37 (1.720) 4.88 (1.598)
No arguments 2.71 (1.624) 3.53 (1.224) 3.16 (1.468)
Total 3.42 (2.057) 3.49 (1.643)
p-value product type .447
p-value type of arguments .000
Table 12: Main results ANOVA attribution of responsibility
DV: Satisfaction with defensive response
Condition Hedonic Utilitarian Total
M M M
Reviewer arguments 4.80 (.469) 3.44 (.382) 4.07 (1.723)
Product arguments 1.93 (.339) 2.93 (.392) 2.47 (1.295)
No arguments 4.10 (.380) 3.45 (.436) 3.75 (1.574)
Total 3.51 (1.900) 3.25 (1.461)
p-value product type .267
p-value type of arguments .000
Table 13: Main results ANOVA satisfaction with response
Attribution of responsibility Satisfaction with response Type of argument Mean
difference
p-value Mean difference
p-value
Reviewer arg Product arg -2.77 .000 .45 .000
No arg -1.04 .000 .08 .542
Product arg Reviewer arg 2.77 .000 -.45 .000
No arg 1.72 .000 -.37 .000
No arg Reviewer arg 1.04 .000 -.08 .542
Product arg -1.72 .000 .37 .000
31 4.1.2 Interaction effects
The ANOVA tests on both attribution of responsibility and satisfaction with the response show that there is an interaction effect between product type and type of arguments (F (2)=10.44, p<.05 and (F (2)=12.215, p<.05. When looking at the plot for the effect on
attribution of responsibility, it can be seen that the lines for no arguments and reviewer related arguments are somewhat parallel. The product related arguments line, however, shows that the attribution of responsibility is much higher in the hedonic condition than in the utilitarian condition. This means that for hedonic products, the attribution of responsibility is more assigned to the retailer when presented with product related arguments, in comparison with utilitarian products. For reviewer related arguments, the responsibility is less attributed to the retailer in the hedonic condition.
The plot for the interaction effect on satisfaction with the response shows that, for hedonic products, the satisfaction with the response is very high for reviewer related and no arguments compared to product arguments. For utilitarian products, this difference is very small. It seems that the defensive response is evaluated positively when presented with reviewer arguments in the hedonic condition, and negatively for product arguments. In the utilitarian condition however, there is almost no difference between the types of arguments on how the defensive response is evaluated.
32
4.2 Outcomes regression analysis
A regression analysis has been performed to see how the independent variables influence the mediator and the dependent variable. Two analyses have been done, on attribution of
responsibility and on satisfaction with the response. The direct effects of the independent variables on attribution of responsibility and satisfaction with the response respectively were tested first. Then, also interaction effects between the independent variables are added, and lastly the covariables skepticism towards others and towards organizations. All regression models had a significance of .000 and all VIF scores were below 4, so no multicollinearity occurred. The output of the regression analysis can be found in Appendix H.
4.2.1 Regression model with DV = attribution of responsibility
In the first regression model (see table 15), only the independent variables are included. It can be seen that product type does not influence the attribution of responsibility, but the type of arguments does. When there are reviewer arguments presented, the attribution of
responsibility to the retailer goes down compared to the no arguments situation. When
33
Attribution of responsibility
Model 1 2 3 4
Constant -.268 .070 .070 .071
IV1 (product type hedonic) -.022 -.221** -.237** -.238** IV2 (reviewer arguments) -.265*** -.277*** -.281*** -.281*** IV2 (product arguments) .453*** .221** .208** .207** IVD1 X IV2
(Product type hed * reviewer arg)
.019 .029 .028
IVD2 X IV2
(product type hed * product arg)
.391*** .403*** .402***
Control variable 1
(Skepticism toward others)
.102* .101*
Control variable 2 (skepticism toward org)
.017
R2 .402 .461 .478 .477
R2 adjusted .393 .447 .461 .457
Table 15: Regression model attribution of responsibility 4.2.2 Regression model with DV = satisfaction with response
In the first model (table 16), with just the independent variables, only product related arguments have a significant effect on satisfaction with the response. It is a negative effect, when presented with product arguments the response is more negatively evaluated. When the interaction effects are added in model 2, the second interaction term is significant. This means that when the product type is hedonic and the arguments are product related, the satisfaction with the defensive response goes down compared to the utilitarian condition. This suggests that people like the defensive response less when presented with product arguments about the hedonic product. In the third model, attribution of responsibility, the mediating variable in this research, is added as a control variable to see how this affects the dependent variable. If attribution of responsibility is added in the model, this has a significant effect on the
34 has a positive significant effect on satisfaction with the response. This suggests that the less skeptic people are towards organizations, the more positively they evaluate the defensive response from the retailer. It can be seen in table 16 that the R squared goes up when adding the interaction effects to the model, and also when attribution of responsibility is added. Adding the first covariable does not increase the fit of the model to the data, but the second one does.
DV: Satisfaction with response
Model 1 2 3 4 5
Constant .091 .026 .031 .031 .034
IV1 (product type hedonic) .046 .197* .139 .145 .136 IV2 (reviewer arguments) .076 .010 -.062 -.071 -.066 IV2 (product arguments) -.386*** -.144 -.087 -.087 -.090 IVD1 X IV2
(product type hed * reviewer arg)
.108 .113 .114 .105
IVD2 X IV2
(product type hed * product arg)
-.408*** -.306*** -.305** -.305***
Attribution of responsibility -.262*** -.265*** -.267*** Control variable 1
(Skepticism toward others)
.002 -.013
Control variable 2 (skepticism toward org)
.153**
R2 .188 .280 .317 .315 .343
R2 adjusted .175 .261 .295 .289 .314
Table 16: Regression model satisfaction with the response
4.3 Mediation analysis
In this section, the results of the mediation analysis will be presented. The overall model suggests that the effect of the independent variables, product type and type of arguments, on the dependent variable satisfaction with the response runs through the mediating variable attribution of responsibility. The three conditions of Baron & Kenny (1986), as mentioned in chapter three, have been fulfilled (see table 17). There is at least one dummy with a
35 significant effect on the dependent variable. This means that mediation analysis is
appropriate.
Condition Beta-value
IV – Mediator (path a)
Product type (hedonic) -.022 Type of
arguments
Product related .453*** Reviewer related -.265***
IV – DV (path c)
Product type (hedonic) .046 Type of
arguments
Product related -.386*** Reviewer related .076
Mediator – DV (path b) -.480***
Table 17: Results conditions of Baron & Kenny (1986) 4.3.1 Explanation of the model
To test whether mediation occurs, a mediation analysis has been performed with the conceptual diagram (model 9) from Hayes (2013). In the model, only one X variable is allowed. Therefore, the second X variable type of arguments is included as a moderator. Because type of arguments has multiple categories, the dummies are added as the W and Z. The model also tests their effect on the mediator M, besides the interaction effect. A second analysis was done to measure the direct effect of type of arguments, as the X variable, on Y. Model 7 from Hayes (2013) was used this time, because here only one moderator is included as product type has a single dummy. The output of the mediation analysis can be found in Appendix I.
X = Product type
M = attribution of responsibility Y = satisfaction with response W = Reviewer arguments dummy Z = Product arguments dummy
